A HEDONIC PRICING ANALYSIS: EVALUATING PRICES

OF BANGKOK’S NEW CONDOMINIUMS

ALONG BTS SKYTRAIN

BY

MISS TISSANA KULKOSA

AN INDEPENDENT STUDY SUBMITTED IN PARTIAL

FULFILLMENT OF THE REQUIREMENTS FOR

THE DEGREE OF MASTER OF SCIENCE

PROGRAM IN FINANCE (INTERNATIONAL PROGRAM)

FACULTY OF COMMERCE AND ACCOUNTANCY

THAMMASAT UNIVERSITY

ACADEMIC YEAR 2016

COPYRIGHT OF THAMMASAT UNIVERSITY

A HEDONIC PRICING ANALYSIS: EVALUATING PRICES

OF BANGKOK’S NEW CONDOMINIUMS

ALONG BTS SKYTRAIN

BY

MISS TISSANA KULKOSA

AN INDEPENDENT STUDY SUBMITTED IN PARTIAL

FULFILLMENT OF THE REQUIREMENTS FOR

THE DEGREE OF MASTER OF SCIENCE

PROGRAM IN FINANCE (INTERNATIONAL PROGRAM)

FACULTY OF COMMERCE AND ACCOUNTANCY

THAMMASAT UNIVERSITY

ACADEMIC YEAR 2016

COPYRIGHT OF THAMMASAT UNIVERSITY

(1)

Independent Study Title A HEDONIC PRICING ANALYSIS:

EVALUATING PRICES OF BANGKOK’S NEW

CONDOMINIUMS ALONG BTS SKYTRAIN

Author Miss Tissana Kulkosa

Degree Master of Science (Finance)

Major Field/Faculty/University Master of Science Program in Finance

(International Program)

Faculty of Commerce and Accountancy

Thammasat University

Independent Study Advisor Associate Professor Chaiyuth Punyasavatsut, Ph.D.

Academic Year 2016

ABSTRACT

Increasing numbers of condominium are seen in the center of Bangkok

Metropolitan Area since BTS SkyTrain began operating in 1999. Areas along BTS are

popular spots to capture benefits and convenience this rapid transit offers. This study

investigates the willingness to pay for a new condominium located along BTS

SkyTrain within one kilometer walking distance from the stations. Hedonic approach

is applied to study the effects of location, structural, and neighborhood attributes have

on condominium price. Geographically weighted regression (GWR) is conducted to

test for spatial variation that may exist in the data. The results reveal no spatial

dependence problem among the observations. OLS regression is performed to

estimate the relationship condominium attributes have on price. The results reveal six

significant attributes, which are land price, proximity to original BTS stations,

location along BTS Sukhumvit Line, number of storey, car parking, and common fee.

It also shows that the attributes affecting the willingness to pay for condominium vary

in different locations.

Keywords: Hedonic pricing analysis, Condominium price, GWR

(2)

ACKNOWLEDGEMENTS

I would like to express my sincere gratitude to Associate Professor Chaiyuth

Punyasavatsut, Ph.D. for his advice and comments throughout the progress of this

independent study. His kind support is greatly appreciated.

I would also like to thank Assistant Professor Chaiyuth Padungsaksawasdi,

Ph.D. for his useful suggestions in this study.

In addition, I wish to thank all MIF faculties and staffs for their guidance and

generous supports during my time at Thammasat University.

Lastly, special thanks to my family, friends, and classmates for their moral

supports throughout the completion of this Master’s degree.

Miss Tissana Kulkosa

(3)

TABLE OF CONTENTS

Page

ABSTRACT (1)

ACKNOWLEDGEMENTS (2)

LIST OF TABLES (5)

LIST OF FIGURES (6)

CHAPTER 1 INTRODUCTION 1

CHAPTER 2 REVIEW OF LITERATURE 4

CHAPTER 3 THEORETICAL FRAMEWORK 7

CHAPTER 4 RESEARCH METHODOLOGY 10

4.1 Empirical model 10

4.1.1 Ordinary least square (OLS) 10

4.1.2 Geographically weighted regression (GWR) 11

4.2 Data 12

4.2.1 Location attributes 13

4.2.2 Structural attributes 14

4.2.3 Neighborhood attributes 15

CHAPTER 5 EMPIRICAL RESULTS 20

CHAPTER 6 CONCLUSION 25

REFERENCES 27

(4)

APPENDICES

APPENDIX A 31

APPENDIX B 32

APPENDIX C 33

APPENDIX D 34

APPENDIX E 35

APPENDIX F 36

BIOGRAPHY 37

(5)

LIST OF TABLES

Tables Page

4.1 List of variables and expected sign 15

4.2 Descriptive statistics 19

4.3 Descriptive statistics for dummy variables 19

5.1 OLS regression results 20

(6)

LIST OF FIGURES

Figures Page

1.1 Newly completed condominiums in Bangkok 1

1.2 Thailand’s House Price Index 2

3.1 Choice of product attributes explained by bid and offer functions 9

4.1 Locations of condominiums and subareas 16

1

CHAPTER 1

INTRODUCTION

Bangkok’s condominium market has expanded tremendously over the past

five years. Demand for living closer to central Bangkok along BTS SkyTrain and

MRT Subway lines increases as people want to save commuting time and costs.

Figure 1.1 shows that the number of newly completed condominium units in midtown

and suburban area has risen steeply since 2012. A large number of middle class

condominium projects have been launched along the extension of BTS and MRT in

the city fringe of Bangkok. The number of downtown condominium has not moved

much due to very limited room to expand in the city center and land price is extremely

expensive.

Figure 1.1: Newly completed condominiums in Bangkok

Source: CBRE Research, Q1 2016

Thanyalakpark (2011) explained that economic conditions and public transit

systems are the two main factors affecting a soar in demand for condominium and

changes in consumer behavior. Bangkok sees a huge change in development after the

2

operation of BTS SkyTrain and MRT Subway began in 1999 and 2004 respectively.

The developers see the convenience of commuting on these rapid transit systems and

launch the projects in close proximity to the stations in response to consumer’s

demand. Although the prices of new condominium located in these prime areas are

extremely high mainly due to expensive land price but many projects were sold-out

instantly during their pre-sale period. This shows that consumers are willing to pay

high price to gain the benefits of living in the city. O’Sullivan (2011) outlined the

advantages of living in the city include reducing transportation time and costs, and

accessibility to wide range of products and services.

Agency for Real Estate Affairs (AREA) surveyed condominium market at the

end of 2015; there are 1,760 residential projects in Bangkok Metropolitan Region

offering 490,649 units. The percentage of condominiums sold to the total supply is 72

percent compared with 58 percent and 60 percent of single-detached houses and

townhouses respectively. This evidently confirms the increasing popularity trend of

purchasing a condominium in Bangkok. Bank of Thailand publishes House Price

Index (HPI) with an index point of 100 in January 2009 as shown in Figure 1.2.

House prices in Thailand show an increasing trend and condominium price has

increased significantly compared with single-detached house and town house prices.

Figure 1.2: Thailand’s House Price Index

Source: Bank of Thailand

80

100

120

140

160

180

Single-detached house (including land)Town house (including land)

3

Purchasing decision of condominium is influenced by many factors and varies

depending on individual’s prioritized criteria. Condominium stocks available in the

market are also varied in characteristics, which are reflected on their prices

(O’Sullivan, 2011). Generally, characteristics that affect residential house prices are

categorized into location, structural, and neighborhood characteristics (Freeman,

1979). This study will examine the willingness to pay for condominiums based on

attributes that are influential to prices of new condominiums in Bangkok located

along BTS SkyTrain. Hedonic approach will be applied to measure the implicit prices

of these attributes. Hedonic model is based on the idea that a product is a bundle of

attributes and the price reflects implicit prices of these attributes.

Hedonic pricing analyses on willingness to pay for condominiums in Bangkok

are conducted using geographically weighted regression (GWR) and ordinary least

squares (OLS). A number of literatures commented that condominium data contains

spatial dependence problem and OLS will not be efficient and the estimated standard

errors are not consistent, which leads to incorrect test statistics. The study will use

geographically weighted regression (GWR) to account for spatial dependence

problem as introduced by Brunsdon et al. (1996). OLS will also be used and compare

the results with GWR.

This study will examine new condominium data between 2012 and 2015,

providing an update to condominium price evaluation and reflects current market

conditions. It has at least two contributions. First, it uses GWR on condominium

prices in Bangkok, which has never been conducted before. Second, it studies

willingness to pay for condominiums located along BTS SkyTrain, which is an

appealing location for condominium and will be useful for consumers who are

looking to buy condominiums in Bangkok to determine appropriate price of a

condominium along BTS SkyTrain. Developers can also use this study to set a

competitive price as compared to other firms, as well as for strategic planning and

marketing purposes in response to consumer’s needs.

The paper is organized in the following orders: reviews of literatures on

relevant studies, explanation of theoretical framework, research methodology,

empirical results, and conclusion.

4

CHAPTER 2

REVIEW OF LITERATURE

This chapter provides reviews of previous literatures related to condominium

prices and hedonic pricing models used for valuation of house prices.

Rosen (1974) explained that the underlying hypothesis of hedonic is that the

value of differentiated products depends on their utility-bearing attributes or

characteristics. Hedonic prices are the implicit prices of these attributes, depending on

specific amounts of their characteristics, are shown in the observed prices of products.

His paper illustrates that when goods are considered as tied bundles of characteristics

then observed market prices can be comparable and the hedonic price-characteristics

equation is identified by the interaction of demand and supply.

Hill (2013) commented that housing provides an extreme example of

differentiated products as every house is different. Hedonic methods define house

prices as a function of a vector of characteristics and can be applied to construct

quality-adjusted residential property price indexes.

In Thailand, Calhoun (2003) showed the first attempt of using hedonic method

to analyze housing values and to develop House Price Index. Bank of Thailand uses

hedonic price method to estimate Thailand’s HPI, which presents in three categories:

Single-detached house, town house, and condominium (Augsorntung and Sittikul,

2016). Benyasut and Dispadung (2012) supported the use of hedonic HPI as they

claimed that it can better represent economic cycles for calculation of CPI.

Apart from computation of HPI, a few researchers in Thailand used hedonic

price model to determine various factors that affect the price of property. Previous

papers mostly concerned the impact of rapid transit systems (RTS) on land and

property value in Bangkok. Vichiensan (2009) and Malaitham (2013) focused on the

impact of RTS development on land and property prices. It empirically shows that

proximity to urban rail stations attracts condominium residential choice and largely

influences land and property values. Kulkolkarn and Laophairoj (2012) showed

various factors that determine condominium prices including project location,

proximity to BTS and MRT, unit location, unit size, furniture, number of units in a

5

project, land size, construction waiting time, developer reputation, parking space,

common fee, and unit type.

Thamrongsrisook (2011) agreed that distance to RTS affects condominium

price, however, the most influential characteristics on prices are distance to main

streets and public facilities. Samaksamarn (2015) conducted surveyed questionnaires

on the willingness to pay for condominium nearby proposed MRT project in

Changwattana Road. The results reveal main factors influencing the willingness to

pay are high-rise building, main road location, reputation of a developer, occupation,

and income.

Varinpramote (2014) applied hedonic price model to evaluate an impact of

Bangkok’s urban rail system on land value. The study found that land value near

operating urban rail transit stations reveals 10.2% premium on average, whereas,

station announcement and construction have no or small influence on land value.

Keetasin (2015) carried out field survey and concluded that proposed extensions of

rapid transit systems into suburban areas will generate commercial projects and

increase land value. More condominium projects are expected to emerge responding

to higher costs of land price.

Regarding the specification of hedonic model, Butler (1982) explained that the

theory provided no specific functional form. All estimates of hedonic relationship are

more or less misspecified but the impact is somewhat small. Linneman (1980)

suggested that Box-Cox transformation technique can be used to identify appropriate

functional form to estimate the price of urban housing market. Most studies conducted

on Thai data executed hedonic price model using OLS, except for Kulkolkarn and

Laophairoj (2012) who included Box-Cox transformation technique and found that it

the most suitable method compared to other models. Huh and Kwak (1997) studied

the functional form of housing market in Seoul, and the results show Box-Cox

functional form is the best hedonic price model in Seoul. The hedonic model structure

is different from those studied in the US, Japan, and Hong Kong, confirming that

there is no universal functional form.

In searching for suitable functional forms, Du and Mulley (2012) explained

that hedonic models such as those for housing market have problems of spatial

autocorrelation or spatial heterogeneity (spatial dependence) as parts of unexplained

6

variance caused by interdependence between observations as a results of their relative

location in space – spatial dependence. If these effects are not accounted for, it will

give biased estimations of coefficients and spatial error dependence.

Hill (2013) further emphasized on the importance to consider spatial

dependence problem in the estimation of hedonic models for housing market. Widlak

et al. (2015) studied spatial and hedonic analysis of house price in Warsaw and

confirmed that Geographically Weighted Regression (GWR) improves the results as

compared to OLS. Geographically Weighted Regression was introduced by Brunsdon

et al. (1996) to account for spatial dependence problem by adding a point coordinate

to each observation and compute weight matrix based on location in geographical

space. The idea is that observations within the same area have tendency to influence

one another more than observations farther away.

In Thailand, there is only one study by Vichiensan (2009) that implemented

GWR model. He conducted hedonic study of land use and transport interaction in

Bangkok using OLS and GWR, which confirmed that GWR approach outperforms

OLS. A hedonic study of condominium prices using GWR that accounts for spatial

dependence has not been implemented, thus this study provides the first attempt of

using GWR to evaluate prices of condominiums in Bangkok.

7

CHAPTER 3

THEORETICAL FRAMEWORK

This study will analyze the willingness to pay of consumers for purchasing a

condominium in Bangkok based on hedonic model proposed by Rosen (1974). He

explained that the implicit prices of the products can be estimated by regression

analysis. The fundamental assumption of hedonic models is that there is large number

of differentiated products available – spectrum of products – among which choices

can be made. The product is described by 𝑛 objectively measured characteristics.

Thus, 𝒛 = (𝑧1, 𝑧2, … , 𝑧𝑛) represents the vector of coordinates with 𝑧𝑖 measuring the

amount of the ith characteristic contained in each good. Then 𝑝(𝒛) = 𝑝(𝑧1, 𝑧2, … , 𝑧𝑛)

is a hedonic price function that is under the assumption of tied sales of the product,

allowing the comparison between observed market prices of the products as price

differences are recognized as equalizing differences for different attributes they

contain. The implicit price of characteristic 𝑖 or 𝑝𝑖 ≡𝜕𝑝𝜕𝑧𝑖

can be estimated by the

parameter of 𝑧𝑖 from regression analysis. Generally, there is no reason for 𝑝(𝒛) to be

linear as Rosen (1974) explained that the differentiated products are sold in separate

markets that are highly interrelated.

The model shows that the implicit prices of underlying characteristics are

derived from both demand and supply. Suppose that consumers maximize their utility

subject to the nonlinear budget constraint. We get

𝑀𝑎𝑥 𝑈 = 𝑈(𝑥, 𝑧1, 𝑧2, … , 𝑧𝑛) subject to 𝑦 = 𝑝(𝒛) + 𝑥 (1)

where 𝑈 is the utility function from purchasing the product, 𝑥 represents other product

consumed, and 𝑦 represents income. The consumers are required to choose 𝑥 and

(𝑧1, … , 𝑧𝑛) to satisfy the budget and the first order conditions

𝜕𝑝𝜕𝑧1

= 𝑝𝑖 =𝑈𝑧𝑖𝑈𝑥

for 𝑖 = 1, … ,𝑛 (2)

where 𝑈𝑧𝑖𝑈𝑥

is the implicit marginal value a consumer is willing to pay to obtain an

additional one unit of 𝑧𝑖, given the same utility and income level.

Rosen’s hedonic model introduces bid function to tackle spatial equilibrium

issue arisen when conceptualizing the problem of the product with a few underlying

8

characteristics instead of a large number of closely related generic goods. The amount

a consumer is willing to pay for alternative values of (𝑧1, … , 𝑧𝑛) at a given utility

index and income is represented by 𝜃(𝑧𝑖 ;𝑢, 𝑦,𝛼) where 𝛼 is a parameter that differs

from person to person. Thus,

𝑢 = 𝑈(𝑦 − 𝜃, 𝑧𝑖,𝛼) (3)

when 𝑦 and 𝑢 remain constant, the first order condition becomes

−𝑈𝑥𝜃𝑧𝑖 + 𝑈𝑧𝑖 = 0 (4)

𝜃𝑧𝑖 =𝑈𝑧𝑖𝑈𝑥

. (5)

Hence, 𝜃𝑧𝑖 is the marginal rate of substitution between 𝑧𝑖 and money.

From equation (2) and (5), we see that 𝜃𝑧𝑖 = 𝑝𝑖 is where the utility is

maximized. Figure 3.1 illustrates consumer equilibrium, showing two different

consumers; one with value function 𝜃1 obtaining maximum utility at point A, and the

other with 𝜃2, while maximizes his utility at point B. The figure shows different

consumption choice, where the second consumer purchases a product with more

quantity of attribute 𝑧1.

It is also important to look at the production decision to account for

equilibrium of buyers and sellers’ locational decisions. Under constant returns to

scale, the producer maximizes profit

𝜋 = 𝑝(𝒛) − 𝐶(𝒛;𝛽) (6)

where 𝐶(𝒛;𝛽) is the cost per unit and 𝛽 reflects underlying variable in the cost

minimization problem or factor prices and production function parameters. First order

condition gives

𝑝𝑖 = 𝐶𝑧𝑖 (7)

where 𝑝𝑖is the marginal revenue and 𝐶𝑧𝑖 is the marginal cost of production to produce

additional one unit of 𝑧𝑖, given the same profit level. At optimum, marginal revenue

from selling additional attributes equals to their marginal cost of production per unit

sold.

Offer function is incorporated into the model, symmetrically with the bid

function from demand side. ∅(𝑧𝑖;𝜋,𝛽) represents unit prices (per model) that the firm

is willing to take from attribute 𝑧1, given a constant profit. Thus,

9

𝜋 = ∅(𝑧𝑖;𝜋,𝛽) − 𝐶𝑧𝑖 (8)

and when 𝜋 is constant, the first order condition gives

∅𝑧𝑖 = 𝐶𝑧𝑖 (9)

From equation (7) and (9), we obtain ∅𝑧𝑖 = 𝑝𝑖, where producer’s profit is

maximized. As illustrated in Figure 3.1, two plants have different optimum points.

The plant with ∅1 has production and cost conditions that is suitable to produce lesser

amounts of attribute 𝑧1 (point A), while the plant with ∅2 has a comparative

advantage at producing more units of attribute 𝑧1 (point B). Therefore, this evidently

shows that the two firms have distinct value of the parameter 𝛽.

Figure 3.1: Choice of product attributes explained by bid and offer functions.

(attribute 1)

The curve 𝑝(𝒛) represents hedonic price schedule. It reveals prices of the product

where demand equals supply and the market clears. As shown in the above figure, bid

curves kiss offer curves at points A and B. Consumers choose a product containing

the set of attributes where the willingness to pay is tangent to the minimum price they

must pay. Similarly, firms choose to produce goods with the set of attributes where

the willingness to accept is tangent to the maximum price obtainable for that model in

the market.

10

CHAPTER 4

RESEARCH METHODOLOGY

1.1 Empirical model

Housing is considered as differentiated product i.e. houses differ in type, size,

age, style, interior features, utility, and location (O’Sullivan, 2011). Different houses

contain different set of characteristics and have different prices. Hedonic model

framework as proposed by Rosen (1974) and described in the previous chapter, is one

of the methods used to determine the equilibrium price of a house. It is based on the

idea that a house is a bundle of attributes and the price reflects implicit prices of these

attributes. As outlined in literature reviews, hedonic pricing model is proved to

provide good approximations of house prices.

Freeman (1979) examined the application of hedonic approach in estimating

property value. Characteristics of a dwelling can be divided into three categories

namely location, structural, and neighborhood. Location attributes (𝐿)describe the

area where the property is situated, proximity to transportation, scenic views etc.

Structural attributes (𝑆) regard the physical characteristic of the property such as size,

number of room, parking space etc. Neighborhood attributes (𝑁) concern

socioeconomic and physical aspects around the neighborhood. Public amenities,

traffic congestion, air pollution etc. can also be included.

The hedonic price model for residential property can be stated as 𝑝 =

𝑓(𝐿, 𝑆,𝑁). It is important to note that there is no specific model for every case, so

various functional forms can be applied (Butler, 1982). As mentioned in reviews of

literature, spatial dependence problem tends to exist in property price data. This study

will apply geographically weighted regression (GWR) to check for spatial dependence

problem, and then OLS regression model will be conducted.

1.1.1 Geographically Weighted Regression (GWR)

Widlak et al. (2015) explained that using OLS for a city or a region can result

in spatial dependence problem. Geographically weighted regression should be used

with spatially heterogeneous data set in order to analyze the spatial variability of the

coefficients. Brunsdon et al. (1996) introduced the GWR approach that accounts for

11

spatial dependence, which refers to variations in relationships over space and is too

complex to be represented by a simple linear or quadratic trend. They showed an

example of the value added for an additional bedroom, which may be greater in an

area populated by families with children than in an area populated by singles or

elderly couples.

GWR extends the traditional regression framework by allowing coefficient 𝛽

to vary spatially. The process weighs every observation by the matrix 𝑊𝑖 and use OLS

to estimate the local values of the parameters. The function is described as

𝑦𝑖 = 𝛽𝑖0 + 𝛽𝑖1𝑥1𝑖 + 𝛽𝑖2𝑥2𝑖 + ⋯+ 𝛽𝑖𝑛𝑥𝑛𝑖 + 𝜀𝑖 . (10)

Now 𝛽, the estimated parameter, varies in terms of location 𝑖 and can be

estimated by

�̂�𝑖 = (𝑋𝑇𝑊𝑖𝑋)−1𝑋𝑇𝑊𝑖𝑌 (11)

where 𝑊𝑖 represents a matrix of weights specific to location 𝑖, while 𝑋 is the matrix of

attributes and 𝑋𝑇is its transpose. Weighted least squares applies a weighting factor 𝑊𝑖

to each squared difference before minimizing, whereas OLS minimizes the sum of the

squared differences by the coefficient estimates. The essential idea is that for every

location i there presents a “bump of influence” around i in which sampled

observations close to i have more influence in the estimation of i’s parameters than

sampled observations farther away. Hence, weighting an observation according to the

proximity of 𝑖 allows 𝛽 parameters to be estimated that coincides with the claim for

spatial variability of the coefficients when dealing with geographical data (Brunsdon

et al., 1996).

1.1.2 Ordinary Least Squares (OLS)

Regression analysis provides an estimation of the relationship between

dependent and multiple independent variables, predicting unknown quantities from

existing data. Linear regression methods assume that the dependent variable is a linear

function of the independent variables. Least squares method minimizes the sum of

squared differences between the actual data and the predicted values. The linear

function is

𝑦 = 𝛽0 + 𝛽1𝑥1 + 𝛽2𝑥2 + ⋯+ 𝛽𝑛𝑥𝑛 + 𝜀 (12)

12

where dependent variable 𝑦 denotes condominium price and independent variable 𝑥

denotes attributes. Parameter 𝛽 can be obtained by

�̂� = (𝑋𝑇𝑋)−1𝑋𝑇𝑦 (13)

where 𝑋 is the matrix of attributes and 𝑋𝑇is its transpose.

1.2 Data

This study will focus on new condominium launched between 2012 and 2015,

located in Bangkok along BTS Sukhumvit Line and BTS Silom Line within one

kilometer walking distance, which is a distance transit planners can expect people to

walk for rapid-transit service (Walker, 2011). During this peroid, the number of new

condominium units launched has increased significantly. Year 2013 saw the highest

number of new residential launches in which new condominium supply reached

84,250 units, almost double the number of horizontal housing types combined (Bank

of Thailand, 2016). Thanyalakpark (2011) explained that economic conditions and

public transit systems are the two main factors that affect soaring demand for

condominium and changes in consumer behavior.

CBRE Global Research (2015) has surveyed the number of existing

condominium supply in inner city of Bangkok. A large percentage of condominiums

are located in Sukhumvit area where important offices and shopping centers are

located. Sukhumvit is the most established residential area, supported by the core BTS

Sukhumvit Line, with 31 percent of condominium supply in downtown area. The

supply in Silom/Sathorn area, which is supported by BTS Silom Line, accounts for 19

percent. This shows that the most popular condominium locations are along BTS

SkyTrain network.

The dependent variable investigated in this study is represented by 𝒓𝒑𝒔𝒒𝒎,

which is the minimum real price per square meter announced by the developer, not

taking into account any sales campaign or promotion. Independent variables

representing 11 attributes that affect the willingness to pay for condominiums are

included in this study divided into three categories: location, structural, and

neighborhood.

13

1.2.1 Location attributes

The selected area of study is considered as Bangkok’s prime location with

high density of residential and commercial zones. The total of four location attributes

is included in the study.

𝒍𝒂𝒏𝒅𝒑𝒔𝒒𝒎 represents land price of the area where a condominium is located

in and is measured in million Baht per square meter. Land price appraisal of districts

in Bangkok during the period of 2012-2015 can be obtained from the Treasury

Department. This variable is included as land price is the major cost for developing a

condominium project. Scarcity of land for development in inner city of Bangkok

results in very high land price. This variable helps to determine condominium price

based on different proximity to Bangkok’s inner city as the selected area of study

covers inner city and city fringe.

𝒃𝒕𝒔𝒊𝒏 is a dummy variable taking 1 if a condominium is located within 400

meter walking distance to original 23 stations at the time of BTS opening in 1999, and

0 otherwise. BTS SkyTrain is Thailand’s first elevated urban rail system. The network

consists of Sukhumvit line or the dark green line, which runs to the north and the east

from central Bangkok, and Silom line or light green line running to the west and the

south. BTS SkyTrain carries, on average, 637,087 passengers daily (BTS Annual

Report 2015/16). It is the most popular mass transit service as it runs along the main

commercial and residential areas. This variable selects the location within the walking

distance of 400 meters to transit as it is the most commonly cited standard used by

transit planners (Walker, 2011).

𝒃𝒕𝒔𝒔𝒖𝒌 is a dummy variable taking 1 if a condominium is located along BTS

Sukhumvit line, and 0 otherwise. Sukhumvit is the most established residential area in

Bangkok where large portion of condominiums is located (CBRE Global Research,

2015). The area is supported by the core BTS Sukhumvit line.

𝒎𝒓𝒕 is a dummy variable taking 1 if a condominium is located within one

kilometer from MRT Subway station, and 0 otherwise. MRT Subway runs from Bang

Sue to Hua Lamphong and connects to BTS at Chatuchak Park, Sukhumvit and Silom

stations. MRT provides connection from inner city center to city center fringe along

Ratchadapisek Rd where increasing number of office buildings are built. This also

connects to populated residential areas in the northeastern suburbs of Bangkok.

14

1.2.2 Structural attributes

Structural attributes include six variables representing physical characteristics

of a condominium.

𝒅𝒆𝒏𝒔𝒊𝒕𝒚 represents the density of a condominium, measured by total number

of unit/total land size in m2. Higher density means more people will be sharing

facilities and services.

𝒔𝒕𝒐𝒓𝒆𝒚 represents the number of storey of a condominium. Generally, there

are two types of condominium: high-rise (more than eight stories) and low-rise (eight

or less stories). Some people may prefer low-rise condominium as it tends to be

quieter with less density.

𝒅𝒆𝒗𝒂𝒈𝒆 represent the age of a developer, measured in month since the

establishment of a company to the time of condominium launch. Reputation of

developer is one of the attributes consumers consider when purchasing a property.

Consumers tend to put more confidence in reputable and experienced developers for

the quality of condominium built and trust in project completion. Rinchumphu et al.

(2016) studied brand value of property in Bangkok and confirmed that branding has

an influential impact on property price.

𝒄𝒂𝒓 represents the ratio of total number of car parking space to total number

of unit. For instance, if the project has 100 units and 70 parking spaces, this variable

will equal to 0.7. Thailand Building Control Act B.E. 2544 states that there must be

one parking space for condominium unit of 60 m2 or larger or one parking space for

every 120 m2 floor area in the project, whichever is higher. The higher the number

should have a positive impact on the willingness to pay for a condominium.

𝒃𝒆𝒅 is a dummy variable taking 0 if a condominium unit is a studio type, 1 if

it is a bedroom type. The price of a bedroom type condominium is higher than studio

type as people are willing to pay more for function of a room (Li and Brown, 1980).

𝒇𝒖𝒓 is a dummy variable taking 1 if a condominium is fully furnished, and 0

otherwise. Generally, fully furnished condominium should be more expensive.

However, most high class condominiums are unfurnished as high income buyers want

to choose the interior design of a room according to their tastes.

15

1.2.3 Neighborhood attribute

Neighborhood attribute consider physical and social aspects around the

neighborhood.

𝒇𝒆𝒆 represents common fee of a condominium in Baht per square meter.

Common fee is paid by every co-owners to pay for expense such as hiring juristic

person, house keepers, security guards, maintenance of common areas, electricity and

water used in common areas, building insurance, and common facilities like

swimming pool and fitness. The fee is higher if there are more facilities available and

thus the price of a condominium also tends to be higher.

Summary of all variables and expected relationship with condominium price

are outlined in table 4.1.

Table 4.1: List of variables and expected sign

Attribute Definition Variable Expected sign

Dependent variable Real price of condominium per m2 (mn Baht) 𝑟𝑝𝑠𝑞𝑚

Location Land price (mn Baht per m2) 𝑙𝑎𝑛𝑑𝑝𝑠𝑞𝑚 +

Location within 400m walking distance to

original 23 BTS stations (dummy)

𝑏𝑡𝑠𝑖𝑛 +

Location along BTS Sukhumvit Line (dummy)

𝑏𝑡𝑠𝑠𝑢𝑘 +

Connection to MRT (dummy) 𝑚𝑟𝑡 +

Structural Density (number of unit/land size in m2) 𝑑𝑒𝑛𝑠𝑖𝑡𝑦 -

Number of storey of the building 𝑠𝑡𝑜𝑟𝑒𝑦 - , +

Developer’s age (month) 𝑑𝑒𝑣𝑎𝑔𝑒 +

Ratio of parking space to total no. of unit 𝑐𝑎𝑟 +

Bedroom type (dummy) 𝑏𝑒𝑑 +

Furniture included in the room (dummy) 𝑓𝑢𝑟 - , +

Neighborhood Common fee per m2/month (Baht) 𝑓𝑒𝑒 +

This study gathers data information from various housing websites as well as

Google Maps service. First, the list of newly launched condominiums during each

16

year between 2012 and 2015 is obtained from www.kobkid.com. Afterwards,

information of each condominium is gathered from www.thinkofliving.com,

www.homenayoo.com, and www.hipflat.co.th. Then, the location of each

condominium is saved on Google Maps and GPS north and east coordinates are

obtained. The information of condominium are published at the time of its first

openings, hence, it is important to adjust condominium price to real price taking year

2012 as base year. This study contains the total number of 125 observations and will

perform four models based on area of BTS SkyTrain stations, i.e. whole area and

three subareas: North, East, and West-South as shown in Figure 4.1.

Figure 4.1: Locations of condominiums and subareas

Source: Google Maps

17

According to descriptive statistics of whole area model, the average

condominium price per m2 is 113,900 Baht, in which the lowest price is 31,300 Baht

and highest price is 273,300 Baht. Land price per m2 ranges from 6,250 Baht to

125,000 Baht with an average of 42,900 Baht. The percentage of condominiums that

are located within 400 meter walking distance to original BTS stations at the time of

opening and is 41.6 percent. Condominiums that are located along the core Sukhumvit

Line comprises 71.2 percent, while 13.6 percent have connection to MRT within one

kilometer walking distance. The average density in a condominium project is 0.091

unit/m2, with a minimum density of 0.032 unit/m2 and a maximum density of 0.409

unit/m2. The age of developer ranges from 3 to 387 months, which averages at 184.3

months. The ratio of car parking space to total unit is averaged at 0.596, with a

minimum of 0.3 and a maximum of 2.36. Most condominiums in the study are a

bedroom type at 74.4 percent of all observations. Fully-furnished condominiums

make up of 36.8 percent. Common fee per m2 ranges from 24 to 90 Baht, with an

average of 53 Baht.

Data are also grouped into three sub-areas: North, East, and West-South,

number of observation included are 18, 71, and 36 respectively. The average

condominium prices of North and East subareas are higher than whole area average

(136,900 and 115,300 Baht/m2), while the average price of West-South subarea is

lower (99,500 Baht/m2). The average land prices of North and West-South subareas

are higher than overall average (49,900 and 44,600 Baht/m2) whereas East subarea is

lower (40,200 Baht/m2). The percentage of condominiums that are located within 400

meter walking distance to original BTS are 72.22 percent for North subarea, 28.17

percent for East subarea, and 52.78 percent for West-South subarea. All observations

in North and East subareas are located along the core BTS Sukhumvit line and West-

South subarea includes the observations located along Silom line. East subarea

contains the most number of condominiums with MRT connection and made up of

15.49 percent of observation, while North and West-South subareas contain 16.67

percent and 8.33 percent respectively. In terms of density, East subarea has the lowest

density (0.081 unit/m2) and below overall average but North and West-South subareas

are higher (0.096 and 0.109 unit/m2). Likewise for number of storey and developer

age, East subarea shows lower averaged number than North and West-South subareas

18

(𝑠𝑡𝑜𝑟𝑒𝑦: 18.15, 24.17, 24.78 and 𝑑𝑒𝑣𝑎𝑔𝑒 (months): 165.37, 201.67, 213

respectively). The ratio of total number of car parking space to total number of unit is

higher for East subarea than North and West-South subareas (0.625, 0.597, 0.537).

Bedroom type condominium comprises of 88.89 percent for North subarea, 76.06

percent for East subarea, and 63.89 percent for West-South subarea. Fully furnished

units are higher in East subarea, North subarea, and West-South subarea in

descending order (40.85, 33.33, and 30.56 percent). North and East subareas have

higher than overall average common fee (57.17 and 53.56 Baht/month), while West-

South subarea has lower average (49.81 Baht/month).

Summary of descriptive statistics are shown in tables 4.2 and 4.3.

19

Table 4.2: Descriptive statistics

Whole area (n=125) North subarea (n=18) East subarea (n=71) West-South subarea (n=36)

Variable Unit Mean SD Min Max Mean SD Min Max Mean SD Min Max Mean SD Min Max

𝑟𝑝𝑠𝑞𝑚 mn Baht/m2 0.1139 0.0516 0.0313 0.2733 0.1369 0.0435 0.0889 0.2405 0.1153 0.0571 0.0313 0.2733 0.0995 0.0388 0.0444 0.197

𝑙𝑎𝑛𝑑𝑝𝑠𝑞𝑚 mn Baht/m2 0.0429 0.0270 0.00625 0.125 0.0499 0.0103 0.03 0.0625 0.0402 0.0297 0.00625 0.125 0.0446 0.0271 0.018 0.125

𝑑𝑒𝑛𝑠𝑖𝑡𝑦 unit/land area(m2) 0.091 0.0411 0.032 0.409 0.096 0.0294 0.040 0.143 0.081 0.0264 0.032 0.173 0.109 0.0603 0.038 0.409

𝑠𝑡𝑜𝑟𝑒𝑦 number of storey 20.93 13.63 5 56 24.17 13.22 8 45 18.15 13.44 5 55 24.78 13.26 5 56

𝑑𝑒𝑣𝑎𝑔𝑒 month 184.31 124.98 3 387 201.67 125.6 5 374 165.37 122.95 4 378 213 125.38 3 387

𝑐𝑎𝑟 total space/total unit 0.596 0.267 0.3 2.36 0.597 0.194 0.4 1.02 0.625 0.308 0.3 2.36 0.537 0.201 0.35 1.2

𝑓𝑒𝑒 Baht/m2 53 15.13 24 90 57.17 11.69 40 90 53.56 16.34 24 90 49.81 13.80 30 85

Table 4.3: Descriptive statistics for dummy variables Whole area (n=125) North subarea (n=18) East subarea (n=71) West-South subarea (n=36)

Variable Frequency Percentage Frequency Percentage Frequency Percentage Frequency Percentage

𝑏𝑡𝑠𝑖𝑛 0 73 58.40 5 27.78 51 71.83 17 47.22

1 52 41.60 13 72.22 20 28.17 19 52.78

𝑏𝑡𝑠𝑠𝑢𝑘 0 36 28.80 0 0 0 0 36 100

1 89 71.20 18 100 71 100 0 0

𝑚𝑟𝑡 0 108 86.40 15 83.33 60 84.51 33 91.67

1 17 13.60 3 16.67 11 15.49 3 8.33

𝑏𝑒𝑑 0 32 25.60 2 11.11 17 23.94 13 36.11

1 93 74.40 16 88.89 54 76.06 23 63.89

𝑓𝑢𝑟 0 79 63.20 12 66.67 42 59.15 25 69.44

1 46 36.80 6 33.33 29 40.85 11 30.56

20

CHAPTER 5

EMPIRICAL RESULTS

Geographically weighted regression (GWR) is performed on STATA under

two hypotheses. Firstly, does the GWR model describe the data significantly better

than OLS? Secondly, does the set of parameter estimates exhibit significant spatial

variation? The first hypothesis is tested by Monte Carlo simulation and the result

suggests that GWR model is not significantly better for the dataset than OLS (see

Appendix A). The second hypothesis is tested by comparing standard error of

parameter estimates from observed data with those from each run of Monte Carlo

simulation and the results reveal that there is no spatial dependence problem in this

dataset (see Appendix B). These results are different from Vichiensan (2009), which

studied land use and transport interaction in Bangkok and showed that GWR model

describes the data better than OLS. The reason for this difference could arise from the

area where condominium observations are located. The data in Vichiensan’s study is

more disperse, while the data in this study is clustered along BTS SkyTrain stations

and may not contain much data variation spatially. Therefore, this study will focus on

results estimated by OLS as shown in table 5.1.

Table 5.1: OLS regression results Variable Coefficient

Whole area North subarea East subarea West-South subarea

𝑙𝑎𝑛𝑑𝑝𝑠𝑞𝑚 0.600 (0.125)

*** 0.409 (1.581)

0.542 (0.140)

*** 0.835 (0.214)

***

𝑏𝑡𝑠𝑖𝑛 0.0123 (0.005)

** 0.0302 (0.0417)

0.0203 (0.00652)

*** -0.000331 (0.00728)

𝑏𝑡𝑠𝑠𝑢𝑘 0.0174 (0.005)

***

𝑚𝑟𝑡 0.00316 (0.007)

-0.0214 (0.0439)

-0.0122 (0.00772)

0.0348 (0.0134)

**

𝑑𝑒𝑛𝑠𝑖𝑡𝑦 0.0142 (0.0704)

0.0521 (1.059)

-0.0233 (0.142)

0.0525 (0.0644)

21

Table 5.1: OLS regression results (continued) 𝑠𝑡𝑜𝑟𝑒𝑦 0.000288

(0.0002)

-0.000656 (0.0022)

0.000661 (0.00034)

* -0.00041 (0.00035)

𝑑𝑒𝑣𝑎𝑔𝑒 0.0000214 (0.00002)

0.000177 (0.00013)

-0.0000211 (0.00003)

0.0000245 (0.00003)

𝑐𝑎𝑟 0.0286 (0.0113)

** 0.0509 (0.114)

0.0165 (0.0138)

0.0547 (0.0219)

**

𝑏𝑒𝑑 0.00254 (0.0052)

-0.00208 (0.0382)

0.00344 (0.00616)

-0.00953 (0.00727)

𝑓𝑢𝑟 -0.00362 (0.0046)

0.0102 (0.0385)

-0.00635 (0.00527)

0.000721 (0.00687)

𝑓𝑒𝑒 0.00140 (0.0002)

*** 0.000897 (0.00115)

0.00171 (0.00025)

*** 0.00056 (0.00033)

(𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡) -0.0331 (0.0132)

** -0.00959 (0.131)

-0.0189 (0.0192)

0.0733 (0.0201)

𝑛 125 0.800

18 0.128

71 36 Adjusted R2 0.872 0.823 Standard errors in parentheses * p<0.1, ** p<0.05, *** p<0.01

Whole area model

OLS results for whole area model shows significant overall F-test, which

means that all coefficients are not equal to zero, and the adjusted R-squared is 80

percent. The results reveal five significant variables, which include three location

attributes, one structural attribute, and one neighborhood attribute. This study

investigates multicollinearity problem by performing variance inflation factor (VIF).

Multicollinearity problem can increase the variance of estimated coefficients. VIF can

be used to measure the increase of the variance as compared to when the variables are

uncorrelated. VIF equals to one means they are not correlated. For example, a VIF of

1.8 means the variance of a coefficient is 80 percent larger. If VIF ranges between one

and five, the variables are moderately correlated and still considered as acceptable

range. VIF results are all lower than three, indicating that multicollinearity is low (see

Appendix C).

Based on the results, three significant location attributes are land price,

proximity to original BTS stations, and location along BTS Sukhumvit Line. This

supports the results of previous studies (Vichiensan, 2009; Thamrongsrisook, 2011;

22

Kulkolkarn et al., 2012; Malaitham, 2013). Land price and location along BTS

Sukhumvit Line are statistically different from zero at 99 percent confidence level.

The results imply that one Baht increase in land price per m2 will increase

condominium price by approximately 0.6 Baht per m2. Condominiums located in a

popular residential area along BTS Sukhumvit Line enjoy a premium of 17,400 Baht

per m2. Proximity to original BTS stations is statistically different from zero at 95

percent confidence level. It shows that the price per m2 of condominium located

within 400 meters from original BTS stations at its opening is 12,300 Baht higher

than those further away. Connection to MRT shows no relationship with

condominium price. This could result from the observation area that limits to one

kilometer from BTS stations and the sample size of MRT contained in this study is

small.

The results show one structural attribute, which is the ratio of total car parking

space to total number of condominium unit. It is statistically different from zero at 95

percent confidence level and implies that an increase of one of the ratio of parking

space to total unit will increase the price by 28,600 Baht per m2. The sign of

coefficient for car parking space is different from previous study of Kulkolkarn et al.

(2012) that reveal a negative relationship; however, the paper predicted a positive

relationship. The significant positive relationship provides an interesting implication

that although condominiums are located in a proximity to BTS, car usage still has an

impact on willingness to pay for a condominium.

Common fee represents neighborhood attribute of a condominium. The results

show that it is statistically different from zero at 99 percent confidence level. One

Baht increase in common fee per m2 will increase the price of a condominium by

approximately 1,400 Baht per m2. This supports the results from previous studies of

Thamrongsrisook (2011) and Kulkolkarn et al. (2012).

Previous study conducted by Kulkolkarn et al. (2012) revealed a number of

structural attributes that have significant effect on condominium price. This study

indicates that only car park shows significant effect, whereas other attributes

including density, number of storey, developer age, room type, and furniture do not

affect the price. The estimated signs of coefficient of all variables follow prior

prediction, except for density. The predicted sign of coefficient is negative but the

23

results show otherwise. The reason may be that more people want to live in the city

center to get benefits from infrastructures and services, as well as to save

transportation costs and time.

North subarea model

The overall F-test of North model is not significant, which means that we fail

to reject the null hypothesis that all coefficients are equal to zero. The adjusted R-

squared is 12.8 percent. The results show no significant variables. Hence, OLS cannot

be used to explain this model. The reason for this could be that there are only 18

observations included in this model. VIF shows that variables are highly correlated

(see Appendix D).

East subarea model

The results of East subarea show significant overall F-test, which means that

all coefficients are not equal to zero, and the adjusted R-squared is 87.2 percent. VIF

values are lower than four for all variables (see Appendix E). The results reveal four

significant variables. Land price, proximity to original BTS stations, and common fee

are statistically different from zero at 99 percent confidence level, while number of

storey is significant at 90 percent confidence level. An increase of land price by one

Baht per m2 will result in 0.54 Baht per m2 increase in condominium price. The

condominium is located within 400 meter walking distance from original BTS

stations is 20,300 Baht per m2 higher. If the common fee increases by one Baht per

month, the price will increase by 1,710 Baht per m2. Number of storey affects the

price of condominium located along BTS East stations. Additional storey will

increase the price by 661 Baht per m2.

West-South subarea model

The overall F-test of West-South subarea is significant, which means that all

coefficients are not equal to zero, and the adjusted R-squared is 82.3 percent. VIF

values of all variable are lower than five (see Appendix F). There are three significant

variables in this model. Land price is statistically different from zero at 99 percent

confidence level. Connection with MRT and car parking are statistically different

from zero at 95 percent confidence level. The results imply that one Baht per m2

24

increase in land price will increase condominium price by 0.835 Baht per m2. An

increase of one of the ratio of parking space to total unit will increase the price by

54,700 Baht per m2.

25

CHAPTER 6

CONCLUSION

This independent study applies hedonic approach to explore factors that affect

the willingness to pay for condominium located near BTS SkyTrain stations. The

price and characteristics of new condominiums launched between 2012 and 2015 are

collected and categorized into three set of attributes, namely location, structural, and

neighborhood. This study explores four models based on the location of observation,

which are whole area model, North subarea, East subarea, and West-South subarea.

Geographically weighted regression (GWR) is performed and reveals no spatial

dependence problem among the studied condominium prices. Hence, OLS regression

can be used to estimate the relationship that different attributes have on condominium

price.

For whole area model, land price, proximity to original BTS stations, location

along BTS Sukhumvit line, car parking, and common fee show significant

relationship on price. North subarea contains small sample size, which results in high

standard errors. This causes t-test to be unreliable, hence OLS estimations become

biased. The price of condominiums located along BTS East stations are affected by

land price, proximity to original BTS stations, number of storey, and common fee.

West-South subarea condominium price is influenced by land price, connection to

MRT, and car parking.

According to the results, location attributes are the most influential factors on

condominium price. Land price shows significant relationship with price in whole

area model, East and West-South subareas. The willingness to pay is higher for

condominium located along the core BTS Sukhumvit line, especially for

condominiums located within 400 meter walking distance to original BTS Sukhumvit

East stations. Connection to MRT only impacts the price of condominiums located

near West-South subarea or Silom line. For structural attributes, car parking shows

significant relationship in whole area model. However, the results of subarea models

show that it only affects the price of condominium on Silom line. Number of storey

affects the price of condominium located along BTS Sukhumvit East stations.

26

Similarly, common fee representing neighborhood attribute only impacts the

willingness to pay for condominium located near BTS Sukhumvit East stations.

The limitation of this study is that condominium prices included in this study

are collected from the lowest price of condominium projects and not taking into

account any sales promotions that may attract the buyers. Therefore condominium

price that reflects the willingness to pay in this study does not include other incentives

that may affect consumer’s buying decision. Further researches can expand the study

area of condominiums as well as include additional factors; for instance, financial

transaction data such as required down payment amount and marketing campaigns

such as discount or gift package that may affect the buying decision and the

willingness to pay for condominiums.

27

REFERENCES

1. Agency for Real Estate Affairs (2016). This Year’s Houses and Condominiums

Outlook: Prosperous or poor. AREA Press Release No. 35/2559. accessed 7 August

2016, <http://www.area.co.th/thai/area_announce/area_press.php?strquey=press_

announcement1271.htm> (written in Thai)

2. Augsorntung, C. & Sittikul, K. (2016). The Developing of Thailand Residential

Property Price Index. Bank of Thailand Statistics and Information Systems

Department. (written in Thai)

3. Bank of Thailand (2016). Financial Stability Report 2015. accessed 7 August

2016, <https://www.bot.or.th/English/FinancialInstitutions/Publications/FSR_Doc/

FSR2015e.pdf>

4. Benyasut, P. & Dispadung, C. (2012). Residential Property Price: An alternative

approach for house rent in Consumer Price Index of Thailand. Bureau of Trade and

Economic Indices, The Ministry of Commerce, Thailand.

5. BTS Group Holdings PCL (2016). Annual Report 2015/16. accessed 17 August

2016, <http://ir.listedcompany.com/tracker.pl?type=5&id=79845&m=6437b8edc5d3

f70a3f3a973f4f4daaaf7283c3de277007ffe8710d0eca012783&redirect=http%3A%2F

%2Fbts.listedcompany.com%2Fmisc%2Far%2F20160621-bts-ar201516-en-02.pdf>

6. Brunsdon, C., Fotheringham, A. S. & Charlton, M. E. (1996) Geographically

Weighted Regression: A Method for Exploring Spatial Nonstationarity. Geographical

Analysis, 28(4), 281-375.

7. Butler, R. V. (1982) . The specification of hedonic indexes for urban housing.

Land Economics, 58(1), 94-108.

8. Calhoun, C. A. (2003). Property Valuation Models and House Price Indexes for

the Provinces of Thailand: 1992-2000. Housing Finance International, 31-43.

9. CBRE Global Research (2015). Will Prices Rise in Bangkok’s Old

Condominium Building?. CBRE Thailand Research Team. accessed on 17 August

1016, <http://www.cbre.co.th/en/ResearchCentre/Research/Will-Prices-Rise-in-Bangkoks-

Old-Condominium-Buildings>

10. Choiejit, R. & Teungfung, R. (2005). The Influence of Employment Density on

Commuting Patterns in Bangkok: A Case Study of Non-motorized Modes. “Everyday

28

Walking Culture”, the 6th International Conference on Walking in the 21st Century,

31-42.

11. Du, H. and Mulley, C. (2012). Understanding spatial variations in the impact of

accessibility on land value using geographically weighted regression. The Journal of

Transport and Land Use, 5(2), 46-59.

12. Freeman, A. M. (1979). Hedonic Prices, Property Values and Measuring

Environmental Benefits: A Survey of the Issues. Journal of Economics, 81(2), 154-173.

13. Hara, Y., Hiramatsu, A., Honda, R., Sekiyama, M. & Matsuda, H. (2010).

Mixed land-use planning on the periphery of large Asian cities: the case of

Nonthaburi Province, Thailand. Sustainability Science, 5(2), 237-248.

14. Hill, J. (2013). Hedonic Price Indexes for Residential Housing: A Survey,

Evaluation and Taxonomy. Journal of Economic Surveys, 27(5), 879-914.

15. Huh, S. & Kwak, S.J. (1997). The Choice of Functional Form and Variables in

the Hedonic Price Model in Seoul. Urban Studies, 34(7), 989-998.

16. Keetasin, S. (2015). An Analysis of Residential Development Projects Along

Rapid Transit Systems. Real Estate Information Center. (written in Thai)

17. Kulkolkarn, K. & Laophairoj, C. (2012). Attributes Determining Condominium

Prices in Bangkok. Applied Economics Journal, 19(1), 24-45. (written in Thai)

18. Linneman, P. (1980). Some Empirical Results on the Nature of the Hedonic

Price Function for the Urban Housing Market. Journal of Urban Economics, 8(1), 47-68.

19. Li, M. M. & Brown, H. J. (1980). Micro-neighborhood Externalities and

Hedonic Housing Prices. Land Economics, 56(2), 125-141.

20. Liu, L. & Jakus, P.M. (2015). Hedonic Valuation in an Urban High-Rise

Housing Market. Canadian Journal of Agricultural Economics, 63, 259-273.

21. Malaitham, S. (2013). A Study of Urban Rail Transit Development Effects in

Bangkok Metropolitan Region. Graduate School of Engineering, Kyoto University.

22. O’Sullivan, A. (2011). Urban Economics (8th Ed.). New York, McGraw-Hill

Education.

23. Rinchumphu, D., Eves, C. & Susilawati, C. (2016). Brand Value of Property in

Bangkok Metropolitan Region (BMR), Thailand. International Real Estate Review,

16(3), 296-322.

29

24. Rosen, S. (1974). Hedonic Prices and Implicit Markets: Product Differentiation

in Pure Competition. Journal of Political Economy, 82(1), 34-55.

25. Samaksamarn, K. (2015). Willingness to Pay for Condominium Nearby Mass

Rapid Transit Project: Case Study Jangwattana Road. Journal by Faculty of

Commerce and Accountancy, Thammasat University, 146, 25-51. (written in Thai)

26. Thamrongsrisook, C. (2011). The Influence of Rapid Transit Systems on

Condominium Prices in Bangkok: A Hedonic price model approach. KTH

Architecture and Built Environment, Thesis no. 105.

27. Thanyalakpark, K. (2011). Ultimate lifestyle condominium, only steps from BTS

station. Government Housing Bank.

28. Vichiensan, V. (2009). Empirical Study of Land Use / Transport Interaction in

Bangkok. Asian Transportation Research Society, Project No: A08/003.

29. Varinpramote, S. (2014). Bangkok’s Urban Rail System: An impact evaluation.

Erasmus School of Economics.

30. Walker, J. (2011). Basics: walking distance to transit. accessed on 20 October

2016, <http://humantransit.org/2011/04/basics-walking-distance-to-transit.html>

31. Widlak, M., Waszczuk, J. & Olszewski, K. (2015). Spatial and hedonic analysis

of house price dynamics in Warsaw. NBP Working Paper, No. 197.

APPENDICES

31

APPENDIX A

SIGNIFICANCE TEST FOR BANDWIDTH USING

MONTE CARLO SIMULATION

Observed P-Value

0.1509 0.269

32

APPENDIX B

SIGNIFICANCE TESTS FOR SPATIAL DEPENDENCE

Variable Si P-Value 𝑙𝑎𝑛𝑑𝑝𝑠𝑞𝑚 0.0191 0.503 𝑏𝑡𝑠𝑖𝑛 0.0010 0.381 𝑏𝑡𝑠𝑠𝑢𝑘 0.0006 0.649 𝑚𝑟𝑡 0.0011 0.440 𝑑𝑒𝑛𝑠𝑖𝑡𝑦 0.0082 0.329 𝑠𝑡𝑜𝑟𝑒𝑦 0.0001 0.111 𝑑𝑒𝑣𝑎𝑔𝑒 0.0000 0.100 𝑐𝑎𝑟 0.0006 0.894 𝑏𝑒𝑑 0.0003 0.824 𝑓𝑢𝑟 0.0002 0.911 𝑓𝑒𝑒 0.0000 0.580 (𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡) 0.0004 0.942

33

APPENDIX C

TABLE OF VIF AND 1/VIF FOR WHOLE AREA MODEL

Variable VIF 1/VIF 𝑙𝑎𝑛𝑑𝑝𝑠𝑞𝑚 2.65 0.377578 𝑠𝑡𝑜𝑟𝑒𝑦 2.58 0.387524 𝑓𝑒𝑒 2.13 0.468426 𝑐𝑎𝑟 2.12 0. 472414 𝑑𝑒𝑛𝑠𝑖𝑡𝑦 1.95 0. 513359 𝑑𝑒𝑣𝑎𝑔𝑒 1.57 0. 638031 𝑏𝑡𝑠𝑖𝑛 1.41 0. 706982 𝑚𝑟𝑡 1.21 0. 828516 𝑏𝑒𝑑 1.20 0. 834366 𝑏𝑡𝑠𝑠𝑢𝑘 1.20 0. 836159 𝑓𝑢𝑟 1.17 0. 855806 Mean VIF 1.74

34

APPENDIX D

TABLE OF VIF AND 1/VIF FOR NORTH MODEL

Variable VIF 1/VIF 𝑑𝑒𝑛𝑠𝑖𝑡𝑦 9.98 0.100240 𝑠𝑡𝑜𝑟𝑒𝑦 9.00 0.111147 𝑐𝑎𝑟 5.01 0.199717 𝑏𝑡𝑠𝑖𝑛 3.82 0.262105 𝑓𝑢𝑟 3.59 0.278371 𝑚𝑟𝑡 2.92 0.342568 𝑑𝑒𝑣𝑎𝑔𝑒 2.83 0.353182 𝑙𝑎𝑛𝑑𝑝𝑠𝑞𝑚 2.76 0.362617 𝑓𝑒𝑒 1.88 0.532642 𝑏𝑒𝑑 1.58 0.634639 Mean VIF 4.33

35

APPENDIX E

TABLE OF VIF AND 1/VIF FOR EAST STATION MODEL

Variable VIF 1/VIF 𝑠𝑡𝑜𝑟𝑒𝑦 3.55 0. 281514 𝑐𝑎𝑟 3.01 0. 332078 𝑙𝑎𝑛𝑑𝑝𝑠𝑞𝑚 2.90 0. 345174 𝑓𝑒𝑒 2.78 0. 359191 𝑑𝑒𝑛𝑠𝑖𝑡𝑦 2.35 0. 425393 𝑑𝑒𝑣𝑎𝑔𝑒 1.70 0. 588074 𝑏𝑡𝑠𝑖𝑛 1.46 0. 686214 𝑚𝑟𝑡 1.33 0. 754702 𝑏𝑒𝑑 1.17 0. 854098 𝑓𝑢𝑟 1.14 0. 877690 Mean VIF 2.14

36

APPENDIX F

TABLE OF VIF AND 1/VIF FOR WEST-SOUTH

STATION MODEL

Variable VIF 1/VIF 𝑙𝑎𝑛𝑑𝑝𝑠𝑞𝑚 4.41 0. 226685 𝑠𝑡𝑜𝑟𝑒𝑦 2.76 0. 362478 𝑓𝑒𝑒 2.73 0. 365977 𝑐𝑎𝑟 2.55 0. 392611 𝑑𝑒𝑛𝑠𝑖𝑡𝑦 1.98 0. 503905 𝑚𝑟𝑡 1.86 0. 537738 𝑏𝑡𝑠𝑖𝑛 1.78 0. 560994 𝑏𝑒𝑑 1.65 0. 607881 𝑑𝑒𝑣𝑎𝑔𝑒 1.56 0. 639261 𝑓𝑢𝑟 1.35 0. 738452 Mean VIF 2.26

37

BIOGRAPHY

Name Miss Tissana Kulkosa

Date of Birth August 8, 1986

Educational Attainment

2014: M.S. in Finance, Thammasat University

2009: M.Arch., Tsinghua University

2005: BPD (Arch), The University of Melbourne

Work Position Assistant Manager

S.K. Optics

Work Experiences 2016 – present: Assistant Manager

S.K. Optics

2013 – 2014: Graduate Architect

Italian-Thai Development PLC.

2010: Intern Architect

Büro Ole Scheeren, Beijing

2008: Intern Architect

HASSELL Melbourne